Currently, loader jobs are all time-based. So if it depends on the results of another loader job, and the job takes longer than expected, the second job could be acting prematurely on old data. So the only option is to schedule them far apart, and hope the first job doesn't run too long.

This issue can be solved by having a job that is triggered by the success of another job. Since the job is dependent on the data created by another job, it should be triggered any time the first job runs, even if it is kicked off at an arbitrary time.